摘要
研究人体行为图像特征识别问题,在人体动作图像中存在大量的和真实行为无关的掩饰动作特征,掩饰动作的存在会大大干扰人体真实行为特征,造成行为真实特征关联的减弱,形成干扰。传统的关联特征挖掘方法在关联干扰的情况下,很难建立准确的行为特征对应空间,使得关联性发生混乱,造成行为特征挖掘错误。提出了一种采用动作特征人体运行行为挖掘算法。利用黄金分割方法,计算动作特征的权值比重,从而删除冗余特征,为人体行为挖掘提供准确的数据基础。利用非线性分类函数,对人体行为特征进行分类,从而实现人体运动行为的挖掘。实验结果表明,利用改进算法能够有效提高人体运动行为识别的准确性,从而有效地对人体运动行为进行有效识别。
The main research is to identify the characteristics of human action image,and a large number of disguise action features which is irrelevant to the real behavior present in human motion picture,the existence of disguise action would significantly interfere with the characteristics of real human behavior,resulting in weakening the association of the behavioral real characteristics,and forming interference.In order to avoid these defects,a mining algorithm based on motion features of human body operation behavior was put forward.The golden section method was utilized to calculate the weight proportion of motion characteristics,so as to delete the redundant features,and provide accurate data basis for human motion behavior mining.A nonlinear classification function was employed to classify the human behavior characteristics,so as to realize the mining of human movement behavior.The experimental results show that the algorithm presented in this paper can effectively improve the accuracy of human movement behavior mining,thus effectively identify human movement behavior.
出处
《计算机仿真》
CSCD
北大核心
2013年第9期424-427,共4页
Computer Simulation
关键词
运动特征
人体行为
数据挖掘
Movement characteristics
Human behavior
Data mining
